Past Meetup

Getting Hands on with Deep Learning for Images

This Meetup is past

148 people went

Location image of event venue


6:00pm -- Networking and Food
6:30pm -- Introductions, NOVA Data Science Updates, and Announcements
6:45pm -- Presentation and Discussion
8:00pm -- Wrap-up


Deep learning using neural networks is currently one of the most exciting areas of artificial intelligence and data science. Recent advances in GPU hardware and neural network models have allowed researchers to produce state of the art results for applications ranging from image analysis to natural language processing to speech recognition. One of the areas where a huge amount of progress has been made in the last 5 years is detection and classification of objects in images, enabling applications such as diagnosis of disease in MRIs with performance matching that of trained radiologists, or facial recognition software that can power Apple’s secure FaceID.

This session will give a hands-on introduction to neural networks, first moving through general principles of neural networks, and then focusing specifically on convolutional neural networks for image analysis. We’ll work through several image classification and detection tasks using Keras with Tensorflow, demonstrated through Jupyter notebooks that will be available afterwards on GitHub.

We will first cover the basics of building up and training a simple convolutional neural network in Keras from scratch, applying it to a simple dataset like MNIST. We’ll then expand to detection of more interesting image classes in an ImageNet sample using pre-trained networks and transfer learning. Finally, we will look at using pre-built object localization and segmentation frameworks such as Facebook’s Detectron.


Chris Morris, PhD, is a Senior Data Scientist at KeyW, where he works on the application of machine learning to detection of malicious activities in large, cloud-scale cybersecurity data. He is currently the principal investigator for a DARPA program aimed at detecting botnets at internet scales for KeyW. Additionally, he heads an internal research effort on using deep neural networks to perform automated object detection across several imaging modalities. Chris has over 10 years of experience conducting scientific research across several fields for organizations including Johns Hopkins, UCSB, and DARPA.

Capital One
George Washington University Data Science Program